=ni,cu,pd)

PHYSICAL REVIEW B 15 SEPTEMBER 1997-IIVOLUME 56, NUMBER 12

Electronic structure of CeNi 12xPd xSn and LaMSn „M 5Ni,Cu,Pd…

A. SlebarskiInstitute of Physics, University of Silesia, 40-007 Katowice, Poland

A. JezierskiInstitute of Molecular Physics, Polish Academy of Sciences, 60-179 Poznan´, Poland

S. Mahl, M. Neumann, and G. BorstelInstitute of Physics, University of Osnabru¨ck, 49069 Osnabru¨ck, Germany

~Received 2 December 1996; revised manuscript received 18 February 1997!

The electronic structure of CeNi12xPdxSn has been studied by photoemission spectroscopy. CeNiSn be-longs to the class of Kondo insulating materials. The gap formed at the Fermi level is strongly suppressed bysubstituting Pd for Ni. The x-ray photoemission spectroscopy~XPS! valence band spectra can be comparedwith ab initio electronic-structure calculations using the linearized muffin-tin orbital~LMTO! method. Wehave found a small indirect gap and a low density of states at the Fermi energy for CePdSn. The 3d XPSspectra and LMTO calculations indicate a strong hybridization of thef orbitals with conduction band and theinteratomic hybridization which causes the large charge transfer between atoms. We have also observed thecorrelation between the electronic structure near Fermi energy and the crystallographic properties of thealloyed CeNiSn. We also present the electronic structures of LaNiSn, LaCuSn, and LaPdSn. These compoundsare good reference for CeNiSn. At Fermi energy a relatively low density of states is found, for LaCuSn anindirect gap is formed. The metallic samples show a relatively high resistivity at room temperature, the largestfor LaCuSn, which demonstrates the influence of the gap on the electric transport properties.@S0163-1829~97!02936-6#

t1

vim

th

nd

lo

m-4

th-onp

ri

the

-ar

ersrere-thent

turebxed

d-ce-

logofor

leso

uc-

edhec-

r-n in

ofSn

I. INTRODUCTION

A small number of rare earth compounds are knownbehave as semiconductors with very small band gaps ofK.

The 4f compounds in this class always containf elementswhich have unstablef configurations~see Table 1 in Ref. 1!.Transport properties of these materials show Kondo behaat high temperatures and the formation of a gap at low teperatures. Following extensive experimental work onrepresentative Ce intermetallics: CeNiSn,2,3 CeBi4Pt3,4 andCeRhSb,5 the term ‘‘Kondo insulator’’ has come to mean aintermetallic compound which~i! has a nonmagnetic grounstate,~ii ! has an insulating gap, and~iii ! has low-lying exci-tations which exhibit properties of strong correlation, anagous to those found in the heavy-fermion compound.

In addition, in every case there is an isostructural seconductor in which Ce is replaced by a tetravalent nonfelement, when the trivalent 4f analogues are metallic. InRefs. 1, 4, and 6 authors suggest that the hybridization inKondo insulators involves one occupiedf state crossing exactly one half-field conduction band: crossing more thanband leads to a metallic state. Furthermore, there is a gathe top of this conduction band. AtT,TK , whereTK is aKondo temperature, the Kondo resonance forms a filled Blouin zone and a gap of size orderkBTK separates the filledstates from the lowest empty states. In this modeltemperature-dependent properties of these insulators smore easily to be described atT.TK as the thermally in-duced appearance of localizedf electrons, each thermal excitation yields one conduction electron. The materials

560163-1829/97/56~12!/7245~10!/$10.00

o00

or-

e

-

i-

e

eat

l-

eem

e

‘‘dirty metals,’’ in the sense that each local moment scattconduction electrons of the unitary limit. In this temperaturegion the negative temperature coefficient of electricalsistivity results because the resistivity is larger thanMooij limit. 7 The materials which support the developmeof a hybridization gap need to have a simple band strucnear the Fermi level.1 They are, with CeNiSn and CeRhSexceptions, cubic, and their lattice parameters indicate mivalence character for thef element.

The gap is unstable against any change in 4f conductionelectron hybridization caused by alloying, e.g., alloying stuies of CeNiSn reveals that the gap is closed by any replament of about 10% either of the Ni or Ce sublattice.8 Thecharge-neutral substitution of Ce by the nonmagnetic ana~La! is known as a Kondo hole. A finite concentrationKondo holes forming an impurity band in a Kondo insulatgives rise to bound state in the energy gap.9 An impurityband develops with increasing concentration of Kondo hoand can fill the hybridization gap. As a function of Kondhole concentration there is an insulator-metal transition.10

Very recently we have investigated the electronic strture and the magnetic properties of CeNi12xCuxSn,11 themain goal of our systematic investigations of the alloyCeNiSn is to find out the influence of the impurities on tformation of this gap. Now we discuss the electronic struture of CeNi12xPdxSn compounds. Taking into consideation the boundary components of this system, CePdScontrast to CeNiSn is magnetically ordered below Ne´el tem-peratureTN57.5 K ~Ref. 12! and does not form the gap.

In this work we compare to the electronic structurethree La intermetallic alloys: LaNiSn, LaCuSn, and LaPd

7245 © 1997 The American Physical Society

tru

oio

tInal

om

mspid

nionth

Inat

a-

ncedy-

es-nhis

gh-

y

pertheonic

e

7246 56A. SLEBARSKI et al.

for the following reasons. These compounds seem to begood reference samples for a comparison with the isosttural compounds which contain rare earth~RE! elements inthe crystallographic position of La. Second, the gapCeNiSn was found to close with about 10% La substitutfor Ce ~Ref. 13! and is also smeared with increasingx inCeNi12xMxSn whereM 5 Cu or Pd.

The valence electron configuration for La is assumedbe 4f 05d16s2, but La often behaves quite abnormally.particular the high temperature resistivity of La intermetlics is much larger than for the isoelectronicd1s2 systemswith Sc, Y, and Lu. It is as large as the one of the Ce hmologous. This is true for all compounds whose high teperature resistivity has been measured so far@LaAl 2,14

LaRh2,14 LaPd3,15 LaCu6,16 and LaGa6 ~Ref. 17!#. Also thesusceptibility of the La compounds tends to be strongly teperature dependent and relatively large. The electroniccific heat coefficients of the La compounds is often conserably larger than the one of Sc, Y, or Lu.

Finally, Lanthanum alloys have a large volume of the ucell compared with that derived from a linear extrapolatiplotting the correspondent lattice constants againstatomic number of different trivalent RE elements.18 Thevalue of the trivalent ionic radius of La may be too large.fact La would be able to be in a slightly mixed valence stbetween 2 and 3, in most intermetallic compounds or asimpurity @e.g., La:YNi2 ~Ref. 18!#. As a possible explana

FIG. 1. Ce 3d XPS spectra of CeNi12xPdxSn.

hec-

fn

o

-

--

-e--

t

e

en

tion, proposed previously in Refs. 14, 19, and 18, a valeof less than three assumed for La describes the thermonamical properties of La intermetallics.

Recently, in a study of electrical properties of some La3Mbinary phases, whereM represents a transition metal, Gradet al.20 attributed theT2 dependence of the electrical resitivity in the low temperature region and the saturatioreached at high temperatures to spin fluctuation effects. Tmodel, however, is not able to explain the abnormally hitemperature resistivity of LaAl2 above the room temperature.

Very recently the ternary compounds of the LaMSn-typewere investigated.21 The electrical resistivities of LaMSnwith M5 Rh, Ir, Ni, Pd, Pt, Cu, Ag, and Au are again verlarge at room temperature and of the order of 150mV cm forIr and Rh, 300mV cm for Ni, Pt, and Pd, and 400mV cmfor Ag, Au, and Cu.

If taken together these anomalies suggest that the proLa intermetallics seem to be a questionable reference forseparation, e.g., of the resistivity anomally from the phonscattering in the similar Ce intermetallics. The electron

FIG. 2. Magnetic susceptibility multiplied by temperatur(x•T) as a function of temperature for the CeNi12xCuxSn interme-tallics. The ground state is nonmagnetic whenxÞ1 ~the groundstate of CeCuSn is antiferromagnetic!. The effective magnetic mo-mentmeff is proportional to (x•T) 1/2 and is slightly reduced fromthe full 4f 1 Hund’s rule ground state moment of 2.54mB . AtT5300 K meff is about 2.2mB .

a

caoe-oe-foC

lohio

ofghthowngli-

CActralowd

thetcted

ghtom

lf-

n

pro-

repin-

lcu-

0.

56 7247ELECTRONIC STRUCTURE OF CeNi12xPdxSn AND . . .

structure near the Fermi level determines several abnormties observed in Ce as well as in La intermetallics.

Now we present the valence band XPS spectra whichbe compared with calculations of the electronic structurethe valence bands by the self-consistent spin-polarized linmuffin-tin orbital ~LMTO! method. We present the electronic structure of the alloyed CeNiSn by Pd andLaMSn with M5Ni, Cu, Pd. The calculations show the dstructive influence of the substituted element on the gapmation at the Fermi energy. We have also analyzed thecore-level x-ray photoemission spectroscopy~XPS! spectrawhich suggest the high hybridization effects between thecalizedf states and the conduction one. We also confirm tCeNiSn is a mixed-valence compound with the occupatnumber off level nf50.95. We found thatnf is closed to 1when Ni is replaced by Pd.

II. EXPERIMENT

The CeNi12xPdxSn samples prepared by arc meltingthe constituent metals in a cooled copper crucible in a hipurity argon atmosphere, remelted several times andannealed at 800 °C for 1 week. Our room-temperature pder x-ray diffraction studies with a use of the SiemeD-5000 diffractometer show that the compounds are sinphase materials crystallizing in the orthorhombic TiNiS

FIG. 3. Ce 4d XPS spectra of CeNi12xPdxSn.

li-

nfar

f

r-e

-atn

-en-

se-

type structure~space groupPnma! except LaCuSn with thehexagonal structure of the LiGaGe-type~space groupP63mc).

The XPS spectra obtained with monochromatized AlKaradiation at room temperature using a PHI 5600ci ESspectrometer. Total energy resolution of the electron spewere about 0.4 eV. All spectra measured in vacua be6310210 Torr. Calibration of the spectra were performeaccording to Ref. 22. Binding energies referenced toFermi level (eF 5 0!, the 4f levels of gold were found a84.0 eV and the observed energy spread of electrons deteat eF was smaller than 0.4 eV. In the XPS spectra a sliamount of oxygen was detected, which mainly comes frthe impurity phase Ce2O3.23

The electronic structures were studied by the seconsistent tight binding linearized muffin-tin orbital~TB-LMTO! method24,25 within the atomic sphere approximatio~ASA! and the local spin density~LSD! approximation. Theexchange correlation potential was assumed in the formposed by von Barth–Hedin26 and Langreth-Mehl-Hu~LMH !corrections27 were included. The electronic structures wecomputed for the experimental lattice parameters. The sorbit coupling was included for the valence electrons.28 Theband structures of four-component compounds were calated for the supercell Ce8Ni 7PdSn8. The number ofkpoints in the irreducible part of the Brillouin zone was 15

FIG. 4. The valence XPS spectra of CeNi12xPdxSn.

on

th

nn

a-t-

ioin

of

e

-in

ofofunddo

4a

ag-

aprge-

oftal


III. RESULTS AND DISCUSSION

A. The alloyed CeNiSn

The destructive influence of alloying on the gap formatiin CeNiSn is indirectly visible in the Ce 3d XPS spectra.Recently we have presented these spectra forCeNi12xCuxSn system.11 The 3d spin-orbit-split componentsshow in the core Ce XPS spectra in CeNiSn an additiostructure at higher energy side with an energy separatio13 eV in respect to the main 3d5/2f

1 and 3d3/2f1 peaks,

which can be interpreted as a contribution of the 4f 0 con-figuration to the Ce ground state.30 The 3d94f 0 componentsfor both 3d multiplets are clear evidence of the mixed vlence in Ce compounds, they are, however, not present in3d XPS spectra ofg2Ce.29 Based on the GunnarssonSchonhammer calculations30,31 the intensity ratiof 0/( f 01 f 11 f 2), which should be directly related to thefoccupation in the final state indicates anf occupationnf'0.95 for CeNiSn. This 3d9f 0 component for 3d multip-lets has not observed in the Ce 3d XPS spectra of the alloysCeNi12xCuxSn whenx.0.02. Thef level occupancy num-ber smaller than 1 is characteristic for the Ce hybridizatgap materials. The 3d f0 satellite has been observed too

FIG. 5. The total and partial density of statesCeNi0.875Pd0.125Sn. The total and partial DOS are in states/~eVcell! and in states/~eV atom!, respectively. The position of theFermi energy is atE50 eV.

e

alof

he

n

CeNiSn when Ce with unstablef shell is substituted by tet-ravalent Zr even for 10% of Zr concentration.11

Figure 1 presents the Ce 3d XPS spectra ofCeNi12xPdxSn. We first of all consider the componentsthese spectra due to the 3d f0 final states. But in Fig. 1 the3d3/2f

0satellites can only be estimated accurately, the 3d5/2f0

one overlaps the 3d3/2f2 peak. For the peak separation w

used the Doniach-Sˇunjic theory.32 The 3d f0 satellite is dis-tinct only for CeNiSn and with alloying disappeares. Alloying with Pd disrupts the valence fluctuation of Ce ions,consequence of the presence of stable 4f shell, the gap is notevidently visible. A weak shoulder in the high energy tailthe spectrum at about 915 eV is probably due to the tailthe Ce Auger signal expected at these energies. The grostate of the substituted CeNiSn is a singlet due to the Koncoupling ~Fig. 2!. With the substitution, the number ofdelectrons may increase so that the hybridization of thefelectron states with thed states may be weakened. In fact,gradual transition from a valence fluctuation state to a mnetic Kondo state recently has been observed~e.g., inCeNi12xCuxSn with increasingx, Ref. 8!.

In Refs. 6 and 9 authors argue that the hybridization gmaterials are conventional semiconductors, but with a laCoulomb repulsionU amongf electrons. They are the realization of a simple Anderson lattice Hamiltonian.

FIG. 6. The total and partial density of states of CePdSn. ToDOS is in states/~eV cell!, partial DOS is in states/~eV atom!. Theenergy scale is represented relative to the Fermi energy.

n,t t

non-io

a

dth

o

eils

-e

p

e

osi-

of

hattithof6tessery

f Ceider-

n

isvel

e

n


The f conduction state hybridizationD, which describesthe hybridization part of the Anderson impurity Hamiltoniacan be estimated from the shakedown satellites located alow-binding-energy side of the 3d spin-orbit-split compo-nents in Ce XPS spectrum~Fig. 1!. D is defined aspV2rmax,where rmax is the maximum in the density of conductiostates andV is the hybridization matrix. It is possible tdetermine the couplingD energy in terms of the GunnarssoSchonhammer theory. Since the intensity ratr 5 f 2/( f 11 f 2) for Ce has been calculated in Ref. 31 asfunction ofD it is possible to determine the couplingD whenthe intensities of the final 3d f1 and 3d f2 states are measureand separated. The hybridization width estimated bymethod is quite large for CeNiSn~about 115 meV!, the al-loying either with Cu or Pd gives the comparable valuethe couplingD.

Figure 3 presents the Ce 4d XPS spectra. The shape of th4d94 f 1 and 4d94 f 2 region cannot be interpreted in detabecause of the strong 4d9-4 f 1 multiplet interaction effect.The spectra are distorted as the 4f hybridization is increasedwhen one compares them with the 4d XPS spectra of metallic Ce. In addition the Ni 3s peak at 110 eV overlaps the C4d XPS signal. The weak feature of the 4d f final states inFig. 3 is detected for CeNi0.95Pd0.05Sn at energy 124 eV. Theseparation of this satellite in reference to the main 4d f1 XPSsignal in the alloyed sample is comparable with energy se

FIG. 7. Valence band XPS spectra of LaNiSn, LaCuSn, aLaPdSn.

he

is

f

a-

ration of 3d f0 and 3d f1 peaks detected in mixed-valencCeNiSn. Thus, the peak at 124 eV in the Ce 4d XPS spec-trum of alloyed CeNiSn can be attributed to a 4d94f 0 finalstate and indicate the valence fluctuation in Ce atomic ptions.

Figure 4 compares the XPS valence bandsCeNi12xPdxSn intermetallics. The 3d Ni peaks are located 1eV below the Fermi energy, we come to the conclusion tthose states are mostly occupied. The 4d Pd states are abou4 eV beloweF . The XP valence spectra are comparable wthe results of our calculations of the electronic structureCeNiSn, CeNi0.875Pd0.125Sn, and CePdSn. In Figs. 5 andwe present the plots of the total and partial density of sta~DOS! of CeNi0.875Pd0.125Sn and CePdSn. The distributionof Ce DOS for both compounds are nearly the same and vsimilar to Ce DOS recently calculated for CeNiSn.11 Weconcluded that the alloying does not change the shape osubbands, moreover, we have accepted many-body consations, which show that it is not thef level but a renormal-ized f level which strongly hybridizes with a conductioband at the Fermi level. For CeNiSn the total DOS ateF iszero (V-shape gap at Fermi level!,11 Pd when it is substitutedby Ni positions, only slightly enhances this value, which0.36 states/eV atom for CePdSn. The DOS at the Fermi leis three times larger when Ni is replaced by Cu.11 There is acorrelation between DOS at the Fermi level and the N´eltemperature, which is bigger for CeCuSn (TN58.6 K! thanfor CePdSn (TN57.5 K!.

dFIG. 8. Total and partial density of states for LaNiSn.

in

apigtbytic

hth

uce

iSte

iz.eagitn

theies

ndom

doera-

tlyin-rin

ofof

1.5hendn bela-nd

Sn,


The energy bands in the irreducible part of the Brillouzone show the indirect gap ateF for CeNiSn,11 but forCePdSn the bands cross the Fermi level only inG2X andG2Y directions. The tendency to form a hybridization gin either a Pd or Cu compound may be related to the hvalue of resistivity~about 500mV cm) measured for both aroom temperature.33 This gap, however, can be destroyedthe onset of magnetic order. Alternatively, antiferromagnecorrelations may rise to magnetic gaps in the Fermi surfawhich increase the resistivity down to the Ne´el temperature.The DOS shown in Fig. 6 appear quite similar shape to tof CePdSn, very recently presented in Ref. 34 except forenergies close to the Fermi level. We obtained a msmaller value of the density of states at this level than in R34.

Our calculations predict in the CeMSn-type intermetallicsa strong intra-atomic hybridization of Ce-4f orbitals with theconduction band states, which reveal a small gap in CeNThis hybridization points to the possibility of intermediavalence of Ce in CeNiSn.

There has also been observed the interatomic hybridtion which makes a charge transfer between the bandsaddition the alloying deforms the distribution of the chargmainly on Ni, it causes Ni atomic positions to not be equivlent in the unit cell. When Ni is substituted by Cu, the chartransfer is mainly between Ce and Sn, alloying CeNiSn wPd makes a preference to strong charge transfer betweethe transition elements excluding Ce which only shows

FIG. 9. Total and partial density of states for LaCuSn.

h

ce,

atehf.

n.

a-In,-ehall

a

slight change of the total charge. We have noticed that ind bands either Cu or Pd is located at similar binding energof order 3.5 eV with respect toeF and are nearly filled. TheXPS spectra indicate that cerium ions in CeCuSn aCePdSn are in a trivalent or nearly trivalent state at rotemperature. As mentioned before, the 4f compounds in theclass of the ‘‘Kondo insulators’’ always contain the 4f ele-ment having the anomalous valence state, therefore wenot expect the semiconducting properties at the low temptures. The lower DOS ateF in CePdSn in comparison withthe one of CeCuSn could be partly related to a slighsmaller unit cell volume of the Pd alloy. It suggests anverse Brillouin zone relation; for a Pd alloy slightly biggethan for Cu, so the position of the Fermi level is differentboth cases.

B. Electronic structure of LaMSn „M 5Ni, Cu, Pd…

In Fig. 7 a comparison of the XPS valence bandsLaNiSn, LaCuSn, and LaPdSn is shown. A low densitystates at Fermi level is observed. The maxima of thed statesof Ni, Cu, and Pd are located in the spectrum at abouteV, 3.5 eV, and 4 eV, respectively. The distributions of ts,p-like conduction states are clearly visible for LaCuSn aLaPdSn near Fermi energy. The XPS valence bands cacompared with the results of the electronic structure calcutions. Figures 8, 9, and 10 show the calculated total apartial densities of states for LaNiSn, LaCuSn, and LaPd

FIG. 10. Total and partial density of states for LaPdSn.

n


FIG. 11. TB LMTO band structure of LaNiSnalong various symmetry directions.

FIG. 12. TB LMTO band structure of LaCuSalong various symmetry directions.

n


FIG. 13. TB LMTO band structure of LaPdSalong various symmetry directions.

vavxy

1S

eta

c--

thw

exs

uc-

nicpa-

f

si-lberice

rge

respectively. The weak peak at about 6 eV in LaNiSnlence XPS band can be attributed to Sn states, howesome contribution of a Ni satellite cannot be excluded. Ogen contaminations are kept as low as possible. Figures12, and 13 present the band structures of LaNiSn, LaCuand LaPdSn, respectively, plotted along various symmdirections. The energy bands of LaNiSn and LaPdSnsimilar, the indirect gap is formed only inT - Y andS - Rdirections in the Brillouin zone. The high symmetry diretions are occupied ateF . The isostructural Ce-analogous intermetallic compounds evidently form the narrow gap atFermi level in CeNiSn and the indirect gap with very loDOS ateF in CePdSn.

LaCuSn, however, forms the strongly anisotropic gap,cluding theG2A andG2M directions for which the band

-er,-1,n,ryre

e

-

cross the Fermi level. The DOS ateF is for LaCuSn verysmall. Quite different behavior shows the bands of isostrtural CeCuSn,11 the gap is not formed in this case.

In Table I we present some more results of the electrostructure calculations together with experimental latticerameters and residual resistivityr0 and resistivityr(T5300K! - r0.

The f shell of La~Table I! has a fractional occupation othe order of 0.3 due to the charge transfer between thes, d,and f shells. This transfer occurs mainly between the trantion elements and Sn~Table I!. In LaCuSn some additionacharge transfer from La to Cu is seen. An occupation numnf Þ0 is also supported by a consideration of the lattconstants. Plotting the unit volume of, e.g., REPdSn~Ref.12! against the atomic number of the RE’s, La shows a la

d the

52.372

3

TABLE I. The crystallographic properties, the results of TB LMTO band structure calculations, anresistivity of LaNiSn, LaCuSn, and LaPdSn, respectively.

LaNiSn LaCuSn LaPdSnStructure type e-TiNiSi LiGaGe e-TiNiSiSpace group Pnma P63mc PnmaLattice parametersa, b, c in Å 7.696, 4.682, 7.606 4.568, 8.224 7.649, 4.729, 8.0WS radii SLa , SM , SSn in a.u. 3.814, 2.558, 3.366 3.766, 3.094, 3.323 3.860, 2.776, 3Total DOS ateF in states/eV at 1.49 0.05 0.95Charge transferDQ in La, M , Sn 0.09, -0.32, 0.23 -0.25, 0.36, -0.11 0.11, -0.34, 0.2nf occupation number for La 0.39 0.28 0.35Residual resistivityr0 in mV cm ~Ref. 21! 65 285 17Resistivityr~300 K!-r0 in mV cm ~Ref. 21! 300 300 180

tsis

ur

veTe

lu

fr

retl

t

isvin

ato

elsr,

tef

ale

ithdo

aenisathar

oa-w

hera

co-ostor-od-ns-alg

eenS

dsns

m-ointer-orted.

of

urend-

sti-the

ionsvelds

p atvi-n-

or

the

r of


deviation from the linear slope for the trivalent RE elemenThis deviation is of the order of 30% of the one whichobserved for divalent Eu.

As can be seen from Figs. 7 to 10, the band structcalculations for the ordered LaMSn-type intermetallics are ingood agreement with the experimental results, howethere are some questions which are not easy to answer.electronic structures, as determined experimentally or thretically, indicate for the La compounds no abnormal vafor the DOS ateF for the rare earth intermetallics~excludingLaCuSn!. This would predict a rather good conductivity othese alloys, however, the measured resistivity and thesidual resistivity of the analyzed La intermetallics are extmaly large ~Table I!. Moreover, LaCuSn with the indirecgap and a very low DOS ateF shows the largest residuaresistivity r0 and linear temperature dependence ofr, whatsuggests a much larger resistivity at temperatures aboveroom temperature.

Let us consider the mixed valence of La in LaMSn, thenwe would expect a similar behavior of resistivity as itobserved in a small number of rare earth compounds haan anomalous 4f valence state and a very small gap, oftypical order of 100 K.1 In these materials the gap appearsoriginate from a hybridization between the localized f levand the conduction band. The LaMSn compounds, howevedo not behave as semiconductors, although for LaCuSngap ateF is visible in the calculations. The curvature of thtemperature dependence of the resistivity is characteristicthe 4f compounds withf instabilities, and the huge residuresistivity r0 of good quality samples is influenced by thgap. It is well known that doping of CeRhSb or CeNiSn wLa suppresses the gap at low temperatures, however,not significantly change the high temperature resistivity.35,13

Such a large increase in resistivity observed in antiferromnets like CeCuSn or CePdSn may be related to the tendto form a hybridization gap at low temperatures, whichdestroyed by the onset of magnetic order. The antiferromnetic correlations may give rise to a magnetic gap inFermi surface, which would increase the resistivity. Thisgument would not be reasonable in the case of LaMSn com-pounds with a stable or even unstable 4f configuration in theLa ions. From literature we know that the binary phasesthe La3M -type whereM is a transition metal show a saturtion effect and aT2 dependence of the resistivity in the lotemperature region, attributed to spin fluctuations.20 This ef-fect would be able to reduce the charge gap.

Assuming the La abnormalities observed in tLaMSn-type compounds, the large resistivity, the tempe

, J

.

e

r,heo-e

e--

he

g

he

or

es

g-cy

g-e-

f

-

ture dependent susceptibility, and the slightly enhancedefficient g in the specific heat seem to be common for mof the La intermetallic compounds and alloys. These abnmalities have been interpreted on the basis of different mels. For example, the magnetic susceptibility and the traport properties of LaAg show up to 300 K an abnormbehavior, in contradiction to other nonmagnetic RE Acompounds,36 due to a calculated high DOS neareF ,whereas a very similar behavior of the resistivities has bfound in LaXSn compounds with a small or very small DOat eF .

We can cite more examples of different La compoundescribed in the literature with different model descriptiowhich are rather contradictory.

Our calculations of the band structures of some La copounds suggest a valence of less than three for La. This pof view seems to be consistent for a wide group of La intmetallics and explains not only the magnetic and transpproperties but also the volume abnormalities, often observ

IV. CONCLUSIONS

In this paper, we investigated the electronic structureCeNiSn when Ni is replaced by Pd, and LaMSn whereM5Ni, Cu, and Pd. The XPS spectra and band structcalculations are presented for La intermetallics aCeNi12xPdxSn. The Ce 3d XPS spectra indicated the mixedvalence Ce in CeNiSn and trivalent Ce, when Ni is subtuted by Pd. We observed very good agreement betweenXPS valence bands and the calculated one. Our calculatrevealed an indirect semiconducting gap at the Fermi lefor CeNiSn and LaCuSn, while the other compoun~CePdSn, LaNiSn, and LaPdSn! form the gap only in somedirections in the Brillouin zone. The LaMSn compounds didnot behave as semiconductors, although the indirect gaeF is visible in the calculations. The huge residual resistities of the LaMSn intermetallic compounds seem to be ifluenced by the gap.

ACKNOWLEDGMENTS

This work was partly supported by the Foundation fPolish Science~Project: SUBIN No. 10/95!, by the Deut-scher Akademischer Austausch Dienst~DAAD !, and by theDeutsche Forschungsgemeinschaft~DFG!. One of us~A.J.!thanks the State Committee for Scientific Research forfinancial support~Project No. 2p30200507!. The calculationswere made in the Supercomputing and Networking CentePoznan´.

1Z. Fisk, P. C. Canfield, J. D. Thompson, and M. F. HundleyAlloys Compd.181, 369 ~1992!.

2T. Takabatakeet al., Jpn. J. Appl. Phys., Part 126, 547 ~1987!.3T. Takabatakeet al., Phys. Rev. B41, 9607~1990!.4M. F. Hundleyet al., Phys. Rev. B42, 6842~1990!.5S. K. Malik and D. T. Adroja, Phys. Rev. B43, 6277~1991!.6G. Aeppli and Z. Fisk, Comments Condens. Matter Phys.16, 155

~1992!.7J. H. Mooij, Phys. Status Solidi A17, 521 ~1973!.

. 8T. Takabatakeet al., J. Magn. Magn. Mater.76/77, 87 ~1988!.9P. Schlottmann, Phys. Rev. B54, 12 324~1996!.

10P. Schlottmann and C. S. Hellberg, J. Appl. Phys.79, 6414~1996!.

11A. Slebarskiet al., Phys. Rev. B54, 13 551~1996!.12S. K. Malik et al., Phys. Rev. B40, 2414~1989!.13F. G. Aliev et al., Physica B163, 3581~1990!.14D. Wohlleben and B. Wittershagen, Adv. Phys.34, 403 ~1985!.15H. Schneider, Ph.D. thesis, Universita¨t zu Koln, 1984.

e

s.

an,


16E. Bruck et al., Z. Phys. B63, 155 ~1986!.17Y. Tagawa, J. Sakurai, and Y. Komura, J. Less-Common M

119, 269 ~1986!.18 A. Slebarski, J. Jelonek, and D. Wohlleben, Z. Phys. B66, 47

~1987!.19A. Slebarski and D. Wohlleben, Z. Phys. B60, 449 ~1985!.20C. S. Garde, J. Ray, and G. Chandra, J. Alloys Compd.198, 165

~1993!.21F. Canepa and S. Cirafici, J. Alloys Compd.232, 71 ~1996!.22Y. Baer, G. Busch, and P. Cohn, Rev. Sci. Instrum.46, 466

~1975!.23G. Nakamotoet al., J. Phys. Soc. Jpn.64, 4834~1995!.24O. K. Anderson, Tight-Binding LMTO ver. 4.7., 1994.25O. K. Andersen, O. Jepsen, and M. Sob, inElectronic Structure

and Its Applications, edited by M. Yussouff~Springer, Berlin,

t.1987!, p. 2.

26U. von Barth and L. Hedin, J. Phys. C5, 1629~1972!.27C. D. Hu and D. C. Langreth, Phys. Scr.32, 391 ~1985!.28B. I. Min and Y. R. Jang, J. Phys.: Condens. Matter3, 5131

~1991!.29E. Wuilloud, H. R. Moser, W. D. Schneider, and Y. Baer, Phy

Rev. B28, 7354~1984!.30O. Gunnarsson and K. Scho¨nhammer, Phys. Rev. B28, 4315

~1983!.31J. C. Fuggleet al., Phys. Rev. B27, 7330~1983!.32S. Doniach and M. S˘unjic, J. Phys. C3, 285 ~1970!.33J. Sakuraiet al., J. Magn. Magn. Mater.90/91, 422 ~1990!.34T. J. Hammond, G. Gehring, M. Suvasini, and W. Temmerm

Phys. Rev. B51, 2994~1995!.35P. C. Canfieldet al., J. Magn. Magn. Mater.108, 217 ~1992!.36F. Merlo and F. Canepa, J. Phys. F17, 2373~1987!.

=ni,cu,pd)

Documents